Graph attention network-based resource assignment in quantum key distribution optical communication systems R. Selvakumar, D. Kanchana, K. M. Dhanalakshmi, M. Thenmozhi, J. Jeneetha Jebanazer Journal of Modern Optics, 2026 The growing need for secure optical communications in the face of quantum computing threats demands advanced solutions for resource allocation in Quantum Key Distribution Optical Networks (QKD-ONs). This paper introduces a novel framework, termed Lightweight Block Encryption-based Multi-Aspect Graph Attention Network with Gorilla Troops Optimizer (LBEMAGAN-GTO), to address routing and resource assignment challenges in the quantum signal channel. The framework combines three core elements: an Ultra Lightweight Block Encryption Algorithm (ULBEA) for efficient and secure quantum key encryption, a Multi-Aspect Graph Attention Network (MAGANet) to enable intelligent routing and resource allocation, and the Gorilla Troops Optimizer (GTO) to dynamically minimize training loss. Simulation results on the UBN24 topology confirm the effectiveness of the approach, yielding a blocking probability of 0.02, throughput of 6.2 Gbit/s, and fast encryption/decryption times of 0.2s and 0.23s. Additionally, the model ensures high resource utilization and favorable QBER, outperforming state-of-the-art methods for quantum-secure optical networks.
Quantum Computing Applications for Geophysical Modeling of Earthquakes and Volcano Eruptions P. Umaeswari, M. Thenmozhi, P. Vinod Kumar, K.N.S.K. Santhosh, J. Srinivasan, Ponmurugan Panneer Selvam Disaster Advances, 2025 The growing complexity of geophysical systems, like earthquakes and volcanic eruptions, requires computational models that can manage enormous, nonlinear and multidimensional datasets in real time. Classical computing methods still yield results but are often not designed to cope with the scales and stochasticity of seismic and volcanic observations, so quantum computing provides a disruptive technology to tackle this issue, enabling geophysical modeling to entirely transform into a capacity to process and analyze complex patterns at massive scales. This study provides an overview of the potentials of various quantum algorithms such as the Variational Quantum Eigensolver (VQE), the Quantum Approximate Optimization Algorithms (QAOA) and quantum-enhanced Monte Carlo simulations to simulate geophysical processes. The results of these models will be of particular relevance to modeling partial differential equations, inverse problems and tasks of uncertainty quantification that describe seismic wave propagation, magma chamber flow and tectonic stress diffusion. We will also discuss how quantum machine learning (QML) models can improve the forecasts of earthquake epicenters, fault detections and eruption forecasts utilizing quantum feature spaces. Further, we will include a discussion of both quantum sensors and edge quantum processors, with attempts for in situ real-time data collection and data processing in hazardous areas.
Taxi Fare Prediction using Random Forest: A Data-Driven Approach based on Pickup/Dropoff Locations and Ride Attributes Sundravadivelu K, Karthigadevi K, Sakthimohan M, Elizabeth Rani G, M. Thenmozhi, Sanjeevi Kumar V Proceedings of the 7th International Conference on Intelligent Sustainable Systems Iciss 2025, 2025 The accurate estimation of taxi fares is critical for enhancing pricing fairness and optimizing operations within the transportation sector. This paper presents a Random Forest-based approach for predicting taxi fares, utilizing key factors such as the pickup and drop-off locations, timestamp of pickup, and passenger count. A sizable dataset of past taxi trip data is used to build the model, which enables the computer to discover intricate relationships and patterns among the many input variables. The efficiency of the model in producing accurate fare estimates is evaluated using common metrics, such as mean absolute error (MAE) and root mean squared error (RMSE). The Random Forest model provides a reliable solution for real-time fare prediction and exhibits considerable accuracy gains when compared to more straightforward techniques. The findings demonstrate how this strategy might increase fare transparency and operational effectiveness for taxi services, which would be advantageous to both passengers and service providers by providing more accurate and up-to-date fare estimates.
Ensemble Based Feature Selection Method for DDoS (EBFM-DDoS) Attack Detection of Healthcare Data in the Cloud Environment A. Somasundaram, S. Devaraju, V.S. Meenakshi, S Jawahar, M. Manimaran, M. Thenmozhi Cybersecurity in Healthcare Applications, 2025 Cloud computing is a novel technological advancement that provides diverse range services to its users. The significant expansion of cloud computing has stimulated researchers to develop new business models by facilitating the exchange of critical computing resources online. Malicious entities used this technology to disrupt cloud services by launching Distributed Denial of Service (DDoS) assaults. These attacks might be a considerable threat in cloud infrastructure as they flood servers with traffic, leading to their incapacitation in healthcare data. The direct consequences of service failure include financial losses and heightened stress levels for professionals, which can be mitigated by detecting DDoS attacks early and preventing them from affecting the system. Unfortunately, detecting DDoS attacks is an exceedingly challenging task. Accordingly, a DDoS defense framework is essential to prevent and detect the attack or to mitigate its effect on healthcare data at the cloud environment using sophisticated learning techniques. For any learning model, collecting the data is a primary stage in which, especially for investigating the packets to identify the normal and the attack, a huge number of packet characteristics, termed as attributes, have to be collected, forming an input dataset. However, not all the characteristics of network data collected will be essential for the underlying study. Also, these irrelevant attributes always stimulate the performances of entire model and lead to misclassification. This chapter introduces the Ensemble-based Feature Selection Framework (EBFM-DDoS) which has the goal of determining relevant features for DDoS detection. This methodology amalgamates a range of feature selection approaches, namely wrapper, filter and embedded methods.
Quantum Enhanced AI to Predict the Traffic Flow Thenmozhi M, Abishake S, Dhayal S R, Rohit Kanna P R 2nd International Conference on Emerging Research in Computational Science Icercs 2024, 2024 This research work proposes a quantum traffic flow prediction system based on the difficulty of traffic congestion in cities. The system uses real-time traffic data, details on weather condition, and the behavioral patterns of drivers flexibly by incorporating machine learning along with quantum algorithms as an integrated approach. The data is collected through different data collection instruments namely, traffic sensors, GPS devices, and weather stations, making the data set holistic for analysis. In this model, the basic traditional algorithms of Long Short- Term Memory (LSTM) and Convolutional Neural Networks (CNN) are applied together with quantum-based optimization approaches including the Quantum Genetic Algorithm (QGA) and Learning Vector Quantization (LVQ). The QGA component is involved in the process of feature selection and modeling parameters, LVQ operates on traffic patterns classification. Together, these quantum-enhanced algorithms increase system efficiency and ensure efficient input-output work, especially when working with extensive and variable data arrays. Hence, the present research captures the flow of the model development process and establishes the evaluation of the current model’s efficiency over the conventional prediction models. Outcomes indicate considerable enhancements in predicting novelties and regulating congestion characteristic of the system’s suitability for smart city structures and urban traffic applications.
Predictive Maintenance of Machine Tools M. Thenmozhi, Kavya A, Vishnu M, Sumaiya Fathima Proceedings of 2024 International Conference on Science Technology Engineering and Management Icstem 2024, 2024 The study takes an empirical approach by testing the proposed predictive maintenance system in a real industrial scenario. To begin, data is collected from diverse sources within the industrial ecosystem, which includes information gathered through system, which combines data acquired from sensors, devices with PLCs (programmable logic controllers), and different communication protocols. This rich and diverse dataset is then processed and made accessible through a Data Analysis Tool. One of the key strengths of this approach is its reliance on Machine Learning, specifically the logistic regression and support vector machine. It is well-suited for predictive maintenance because of its ability to handle complex datasets and relationships among different variables. The paper details the process of applying the logistic regression and support vector machine approach to the collected data, which allows for the prediction of various machine states. This is crucial in foreseeing potential issues before they escalate into significant failures, enabling timely maintenance interventions. Moreover, the research extends its assessment by comparing the outcomes of the Machine Learning approach to results obtained using simulation tools, thus providing a comprehensive an assessment of the system for predictive maintenance.
Sentiment Analysis on Climate Change using Twitter Data M. Thenmozhi, G. Shubigsha, G. Sindhuja, V. Dhinakar Proceedings of the 2nd IEEE International Conference on Networking and Communications 2024 Icnwc 2024, 2024 The objective is to illuminate the nuanced and complex public opinion surrounding climate change conversations on Twitter. The work employs advanced machine learning methods and natural language processing techniques, notably Support Vector Machines (SVM), to conduct a global-scale sentiment analysis using a sizable dataset that was obtained from a reliable third-party source. This study's main goal is to pinpoint the various ways that emotion is expressed in the context of climate change conversation; both positive and negative expressions are taken into account. The authenticity and utility of the selected third-party dataset—obtained through Kaggle and made possible by a Canadian Innovation Foundation JELF grant awarded to Chris Bausch at the University of Waterloo—are critically evaluated. The SVM-based sentiment analysis demonstrates how well the chosen methodology reflects the complexity of sentiment in climate change debates on Twitter, with an exceptional F1 score of 0.70. The implications of the research include communication strategies for legislators, organizations seeking to reach a global audience, and climate change advocates. By utilizing an external dataset and applying the Support Vector Machines algorithm's sentiment analysis, this study advances our comprehension of the intricate relationship between the public's perspective of climate change and social media conversation. To sum up, this study shows the value of SVM in detecting subtleties of sentiment in huge databases, contributing significant new knowledge to the developing subject at the nexus of environmental consciousness and social media dynamics.
The New Paradigm in Education: Visionary Nexus Thenmozhi M, Mahesh V, Aishwarya R, Kaviarasan M, Manoj P D 2024 1st International Conference on Data Computation and Communication Icdcc 2024, 2024 Visionary Nexus is a novel platform designed to bridge the gap between academia and industry, providing students with direct access to industry-leading professionals for mentorship, teaching, and workshops. This paper presents the conceptual framework, design architecture, features, and future scope of Visionary Nexus, highlighting its unique approach to enhancing educational quality. By linking experts with educational institutions worldwide, Visionary Nexus ensures that students are exposed to the latest industry knowledge, preparing them for modern workforce challenges. The platform leverages a feedback-driven rating system, ensuring continuous improvement in teaching standards and expert performance. Additionally, its global outreach promotes diversity and inclusivity in education, equipping students with a broad perspective. As the platform evolves, future integration of AI for expert matching and expansion into emerging disciplines will further revolutionize access to industry expertise. Visionary Nexus is a transformative initiative redefining education, fostering collaboration, and shaping the future of global academic and professional development.
Automated Apple Wax Detection with Convolutional Neural Networks K Sundravadivelu, S Dheenath, M Sakthimohan, G Elizabeth Rani, M. Thenmozhi 8th International Conference on Electronics Communication and Aerospace Technology Iceca 2024 Proceedings, 2024 In this work, the application of wax coatings on fruits is a widespread practice aimed at enhancing visual appeal and shelf life but raises concerns regarding consumer health and transparency. This study introduces an automated approach using IoT and machine learning to detect wax coatings on fruits. Leveraging Convolutional Neural Networks (CNNs), we developed a robust system capable of accurately distinguishing between waxed and unwaxed apples. Our methodology involved collecting a comprehensive dataset, preprocessing images, and training the CNN to classify images based on wax presence. The system demonstrates high accuracy, validated through rigorous testing and comparison with existing methods. This research contributes to enhancing consumer transparency in food safety and sets a foundation for future applications in agricultural technology.
Enhanced Skin Lesion Diagnosis using a Deep Learning Techniques M Sornalakshmi, G Elizabeth Rani, M Sakthimohan, M. Thenmozhi, D Amuthaguka, V Sanjeevi Kumar 8th International Conference on Electronics Communication and Aerospace Technology Iceca 2024 Proceedings, 2024
